The boom in artificial intelligence has brought with it many new capabilities, from speeding up time-consuming processes like drug discovery to ushering in the next generation of search engines. Some companies are already positively taking advantage of this powerful technology, while others are learning that, if left unchecked, AI is capable of delivering unsavory results that can cause harm. Recent examples include AI telling someone to eat rocks, demonstrating discriminatory behavior, and giving incorrect information to a customer that resulted in legal liability.
Aside from human bias and the nascency of the technology, one of the driving forces behind this disparity in quality can be explained by an old computer science phrase: garbage in, garbage out. Information produced by a machine can’t be better than the information that’s going into it. Therefore, organizing and managing data to ensure its quality is one of the most important things a company can do when using AI.
“Bad data—inaccurate and untrustworthy, is in silos everywhere. It can cost companies time, money, and negatively affect brand image. It also impairs predictive and generative AI models, ” said Thomas Wyatt, president of Twilio Segment, a customer data platform that helps businesses collect customer data, create unified customer profiles, and deliver highly personalized customer experiences. “The successful companies of the future will have a strong data foundation to power AI that delivers next generation customer engagement.”
AI Is Only As Good As Your Data
Good data unlocks insights, saves time, and drives growth. Companies that rely on incomplete or inaccurate data may have AI that creates bad outcomes that range from costly mistakes and legal issues to damage to customer loyalty—or worse. This year, regulators fined a major financial services company $136 million for its failure to address data management issues and manage risks.
According to Twilio Segment, the average business has hundreds of tools in its tech stack and roughly 185 apps in its workflow. These disparate data sources make it nearly impossible for an organization to have the cohesive, accurate, and up-to-date information necessary to generate a reliable view of their customer.
One recent study showed that 64% of companies believe that AI will help improve customer relationships. It can, but only if it’s intentionally set up to do so. According to research by Twilio Segment, 89% of survey respondents believe personalization is invaluable to their business’ success in the next three years, but 61% of survey respondents are worried about inaccurate data compromising their personalization efforts. What does it look like for a company to create a system that can merge many different data sources into one customer profile, and how can that help drive positive AI?
Trade Me, an online auction site, had this same question. The company’s goal was to improve marketing campaign performance and run targeted, personalized campaigns without needing to rely heavily on its busy data science team. To do this, Trade Me partnered with Twilio Segment. First Segment unified multiple sources of customer data into complete customer profiles, and then used CustomerAI Predictions, its predictive AI model, to help the marketing team predict which users were most likely to engage.
Now Trade Me can do things like determine which customers are most likely to bid on an auction item and send targeted email campaigns. Ultimately, Trade Me increased open rates by 20%, click-through rates by 10%, and achieved 2 to 3 times better campaign performance and higher return on ad spend.
The Path To Create Good Data
The first step on the journey to good data is to invest in a system, such as a CDP, that can capture and validate customer data from across channels in real time. This data can then be linked to existing data in your data warehouse, and unified into single customer profiles. It’s these unified profiles that can be used by different teams for different purposes, from powering AI models to crafting successful marketing campaigns.
MongoDB, a platform for developing scalable applications, suffered from this problem of siloed customer data. Its data was in CRMs and web analytics tools and other platforms, and each of its teams was using different sourcing, debugging, and maintenance protocols. This resulted in wasted time, data discrepancies, and difficulty getting a comprehensive view of its customers—who happened to be sophisticated and savvy developers.
“Siloed data in a generative AI era is a recipe for disaster. They prevent companies from accessing real-time data to fuel AI-powered engagement,” said Wyatt. “These silos inevitably cause inaccurate or poor-quality data that impair AI models, and generate misleading outputs or 'hallucinations'. These inaccuracies skew the understanding of customer preferences, which erodes customer trust and loyalty over time.”
Once the team at Twilio Segment helps break down these silos by connecting disparate data sources, the next step is to make sure the data is of sufficient quality, which includes compliance with HIPAA and country-specific standards. Twilio Segment Protocols focuses on data governance to ensure that data is clean, accurate, and is shielded by layers of safeguards designed to keep it private and secure. Segment automatically reconciles incorrect data as it's being processed. If the same customer adds ‘USA’ when making a purchase on mobiles, and ‘US’ for another purchase on desktop, Segment knows this is the same shopper. Bad data can also include things like duplicate entries or incomplete fields. Without clean data, there’s always a risk of making decisions based on misleading or inaccurate information, which can lead to huge financial costs. According to recent findings, poor data quality can cost a business around $13 million per year on average.
After unifying its data, MongoDB was able to unlock a foundation of good data capable of generating new insights, opportunities for personalization, and drive engagement. Importantly, it was able to see that when developers got stuck while building their apps they were going to other vendors if they didn’t find helpful information quickly. Before, MongoDB couldn’t see this happening in real time, but now it can use different tools like live chat and pop ups to help deliver the educational content the developer needs, and keep its customers from leaving the site.
“Good data is effectively managed across collection, storage, usage and deletion,” said Wyatt. “Twilio helps companies harness the power of AI by creating a trustworthy data stack. From there, the possibilities are endless.”
How to Unlock the Full Potential of AI
Good data in, good data-driven marketing out. By ensuring that data collection, storage, integration, and analysis are streamlined, and that the data is clean and trustworthy, a company can use that high-quality data to feed its AI and transform it into actionable insights, such as real-time product recommendations.
Many businesses are looking to incorporate AI in order to create highly personalized experiences for consumers. AI can process data far more quickly than a human can in order to do things like generate recommendations, target promotions to individual preferences, or offer custom content. It can also do things like monitor purchasing, so that when a customer actually makes a purchase they are automatically removed from ad targeting in order to avoid wasted ad spend.
One of the newer capabilities at Twilio Segment is CustomerAI, which can unlock personalization, speed up competitive advantage, and drive customer satisfaction using predictive and generative tools like CustomerAI Predictions and Voice Intelligence.
With Predictions, companies can track purchase history, web browsing, and social media and customer service interactions and use it to help predict customer desires and behavior and adjust its actions accordingly and automatically. Voice Intelligence pulls key insights from customer service call center transcripts in order to determine common feedback, complaints, and compliance risks. Companies can also use AI to address other specific needs like reducing churn, increasing engagement or monthly users, or informing product development.
As AI applications continue to advance, smart companies that invest in good data practices and a solid tech stack now, will be equipped to get ahead and be ready for whatever comes next.
